Anyone shot much 35 Mpbs on the 700?

[Alister Chapman]

I think we need to go back to basics here.

A full resolution Y CbCr image is 4:4:4. Subsampling that image to 4:2:2 is a form of compression. Compression is, in it's simplest terms, taking a big thing and making it fit in a smaller space, whether that is through sub sampling, DCT, quantization or any other method, making a big signal smaller is compression.

4:2:2 is the ratio between the amount of data used to portray the Y (4) Cb (2) and Cr (2) signals. So if you use 8Mb/s for the Y then for it to be 4:2:2 you MUST use 4Mb/s for each of the Cb and Cr. If you use anything other than this ratio then it is not 4:2:2 but some other ratio.

The simplest way to achieve this is to use the same compression routine for each channel and simply reduce the sample size by a factor of 2 for Cb and Cr. This is very easy to do. Most 4:2:2 encoding schemes use one encoder to encode the Y and then a second similar encoder running the same coding routine but alternating between one sample of Cb followed by one sample of Cr. In fact a better way to describe this would be 2:1:1 as for every two luma bits of luma data there are one each of Cb and Cr. This gives us the familiar 4:2:2 encoding that we are all familiar with. The compression ratio is by default the same for each channel while the sample size is halved for each of Cb, Cr.

If you compress the chroma (Cb,Cr) by a different ratio than 2:1 then the ratio between Y and Cb, Cr is no longer 2:1 so you can no longer claim the system to be 4:2:2. While it is technically possible to compress the Cb and Cr at a different ratio to the Y channel this is not what is going on with XDCAM and is borne out by the maths.

4:2:2 = (1920x1080 + 960x1080 + 960x1080) x30(fps) x8 (bits) = 995Mb/s
4:2:0 = (1920x1080 + 960x1080) x30(fps) x8(bit) = 746Mb/s

If we divide 995Mb/s by 19.9 (the exact compression ratio given in the Sony document) we get 50Mb/s
If we do the same for 746Mb/s divided by 21.3 (again from Sony document) we get 35Mb/s

I don't think it is a mere coincidence that the maths works so exactly, more to do with the fact that the compression ratio is equal in each channel.


As I said, and on this we agree. Starting with more samples, ie. 422 over 420 should in theory at least give a marginally better end result however this will depend a lot on the quality of the decoder and the image content. In terms of visible artifacts these are most noticeable in the full resolution luma channel and with both XDCAM HD422 and EX this is compressed by pretty much the same amount so there will be little to no difference in most cases. In terms of concatenation there will be very little, if any difference between EX and XD HD422 material. Differences in scene content, lighting and camera settings such as detail, gain and aperture will make a much bigger difference than the difference between 4:2:0 @ 35Mb/s and 4:2:2 @ 50Mb/s. The BBC are currently enforcing the 50Mb/s rule yet don't stipulate gain or detail limits. In the past it was possible to get a dispensation for 35Mb/s with an EX, but last I heard the rules were being much more strictly enforced and it's all but impossible to get permission to use an EX, even for situations where shooting with a larger camera is just about impossible.

As you say it's easy enough to add a NanoFlash to an EX and record at 50Mb/s or better still 100Mb/s to comply with the rules, but to be honest the rules don't make a great deal of sense. The difference between 35Mb/s EX and 50Mb/s XD HD is very small from an artifacts and muti-generation performance point of view. A PDW-700 with a high detail setting or any gain switched in will produce an image that will fall apart quicker in a poor production chain that a well set up EX1. It is arguable that the ultra clean (59db) images from the PMW-350 will hold up better than the noisier pictures from the PDW-700 (54db). My belief is that the BBC have chosen to set the bar at 50Mb/s to prevent or restrict the use of small cameras as they are often operated by people that simply don't know how to use a camera properly in the name of "cost saving". What would make more sense would be to allow 35Mb/s while ensuring that productions use skilled and competent crews.

In the past programmes would be rejected for poor camerawork, dirty lenses and out of focus shots. These days it seems that provided it meets the correct technical specifications pretty much anything goes.

[Simon Wyndham]

I don't know enough about the technicals to chime in regarding the 422 and 420 compression systems. However I do think that the BBC's rules are a little on the silly side. While they may be following the EBU recommendations they should be followed as such. As guidelines.

The trouble with putting a cast iron rule on the type of recording system and size of chips used is that there are barely any cameras that meet the spec!

The Panny 301 falls foul of the rules because of its 1/3" chips. All other large chip Panasonic cameras fail to meet the spec because they don't have full 1920x1080 sensors until you reach the 3000 series.

Of the Sony's you have to go to the 700 and 800 as far as out of the box systems go. It is all well and good that the BBC lay down the law, but they must be prepared to give budgets and pay that allow for people to meet those specifications.

I do think that the rules will be relaxed. Dear old Auntie only has one HD channel, bless her. But as soon as HD spreads more and they have more output they won't be able to use 50Mb/s min spec cameras all the time.

[David Heath]

I think an assumption is being made here, and it's not valid. I agree with the basic numbers you quote above:

Originally Posted by Alister Chapman

4:2:2 = (1920x1080 + 960x1080 + 960x1080) x30(fps) x8 (bits) = 995Mb/s
4:2:0 = (1920x1080 + 960x1080) x30(fps) x8(bit) = 746Mb/s

But let's look at them another way.

For 4:2:2, you can regard the 995Mbs as two separate streams of about 497.5Mbs - one for luminance, one for chrominance. It's possible to then deal with them in separate ways, and that's exactly what happens. But at no stage are the image sizes altered - they remain 1920x1080 (Y) or 960x1080 (U & V).

You are making the assumption that each stream is compressed according to the figure in the Sony document - 19.9 in this case - which implies that after compression we end up with two streams of about 24.9Mbs. Add them together, and hey presto, there's out 50Mbs XDCAM 422 signal.

But you don't need to apply the same compression ratio to each stream, any more than you need to apply the same compression ratio to two still photos. What if we choose 12.44 as the compression factor for luminance, and 49.75 for chrominance? That then gives us two streams post compression of 40Mbs and 10Mbs, add them together and we're back to 50Mbs. Obviously there's an infinite number of pairs of values for compression ratio which give the same total average - I don't know the real ones.

Practically, what this means is that the chrominance image is not just softer than luminance (because of sub-sampling), but will have higher artifacting as well (because of the higher compression ratio).

You can actually see the results of this on broadcast digital TV on low bitrate channels. One effect of a codec breaking down is a smooth gradient taking on a stepped or contoured appearance, and you may have noticed that the effect tends to occur far more on highly coloured gradients than lightly coloured ones. This is a direct result of using a higher compression factor for chrominance than luminance.

If we divide 995Mb/s by 19.9 (the exact compression ratio given in the Sony document) we get 50Mb/s
If we do the same for 746Mb/s divided by 21.3 (again from Sony document) we get 35Mb/s

I don't think it is a mere coincidence that the maths works so exactly, more to do with the fact that the compression ratio is equal in each channel.

No, no co-incidence, but the 19.9 and 21.3 are just averages of input/output data rates using the lum/chrom streams totalled in each case. They don't tell you anything about what's happening individually to luminance compared to chrominance.

A bit like looking at traffic accident statistics. You might be told there are 50 per 100,000 of the population - but that tells you nothing about relative figures for old/young drivers etc.

Let's compare XDCAM EX at 35Mbs to HDV. They are both 4:2:0 systems, but the former is based on 1920x1080, the latter on 1440x1080. If we accept your logic, then to increase the number of pixels from 1440 to 1920 is an increase of 33%. If we apply that pro rata to post compression data rates, we need to increase the 25Mbs by 33% - about 8.5Mbs - to get about 33.5Mbs. So wouldn't you then expect the overall quality of the 35Mbs codec to be about the same as HDV, albeit with a higher horizontal resolution?

In practice, that's not the case - XDCAM EX is substantially better than HDV - and a large reason for that is that you can't simply say "33% more pixels means a 33% increase in coded data rate for the same compression quality". As I've been saying all along.

My belief is that the BBC have chosen to set the bar at 50Mb/s to prevent or restrict the use of small cameras as they are often operated by people that simply don't know how to use a camera properly in the name of "cost saving". What would make more sense would be to allow 35Mb/s while ensuring that productions use skilled and competent crews.

Well, not just the BBC, it's an EBU wide recommendation, and followed by others outside Europe. My own feeling is that here it would make the most sense for the manufacturers to put a 50Mbs mode on the camera than for the broadcasters to relax their preferences. I can't believe it would be technically difficult or expensive, (given the total price of the nanoFlash) far more likely an attempt to force the users to pay more than is really necessary - buy the next model up.

It could be a dangerous game if another manufacturer suddenly came out with a comparable camera that ticked all the boxes. In the meantime, it must be making Convergent Design very happy. :-)

[Simon Wyndham]

I can't believe it would be technically difficult or expensive, (given the total price of the nanoFlash) far more likely an attempt to force the users to pay more than is really necessary - buy the next model up.

In one way yes. But in another Sony could for example release another model that does 100Mb/s and make the 700 a base model.

I think in some respects the 700 has been the victim of the current financial situation. The prices for equipment are now going up again (telling this to people like Producers and Michael Rosenblum however falls on deaf ears). Otherwise it might be possible for the 350 to take over from the DSR series (as intended) and the 700 to take over from the old PDW-350.

Never did get an answer as to why they named the new camera a 350, since people generally just say "350". Now I don't know if they are talking about the disc camera or the SxS camera!

[Alister Chapman]

I guess rather than speculating or hypothesizing the simplest thing to do would be to analyze the streams from an EX and a PDW-700 and see what's actually going on.

Anyone know a good utility that will show the separate Y CbCr bitrates.

However you look at it the compression ratios are remarkably similar and my multi-generation test which is made up of 5 passes using the native codec and shrinking and shifting then stretching and centering the image on each pass gives almost identical results on both XDCAM HD422 and EX. Given this is an unrealistic workflow anyway and a more robust codec should be used through multi-generation post, prores HQ or DNxHD for example, I stand by my claim that there is little to no difference in the robustness of the two formats and that noise/grain and picture content will have a greater bearing on the final image.

Not everyone is following the EBU recommendations and I know there are several influential people within the BBC trying to get acceptance for the EX and in particular the PMW-350. I've seen plenty of my EX material in big budget HD broadcasts alongside HDCAM and it's never looked in any way sub standard. Lost Land of the Volcano was shot using EX's and PDW-700 and I bet no one could tell which was which in the finished programme. Well.... apart from when one of the EX's melted going in to the Volcano.

I'm not saying EX is a replacement for PDW, for a start the workflows are very different. But from the point of view of image degradation through a typical end to end production chain I really don't think there is any significant difference between the two. In fact that's why I like to shoot at 100Mb/s with my NanoFlashes, because at 100Mb/s I am seeing real benefits with both.

[David Heath]

Originally Posted by Alister Chapman

I guess rather than speculating or hypothesizing the simplest thing to do would be to analyze the streams from an EX and a PDW-700 and see what's actually going on.

Yes, I'd love to know the answer. My understanding is that the chroma is definitely being compressed by a different factor to luminance - but I don't know by how much.

However you look at it the compression ratios are remarkably similar.......

I feel you are equating compression ratio (data rate in:data rate out) to compression quality, and even within a given codec I don't believe that's a valid assumption. How well an image compresses is due far less to how many pixels make up the image, but how much information it contains - using the word in the scientific sense. Hence, a 960x1080 image which is uniform blue for the top half, red for the bottom could theoretically be perfectly compressed into only a few bytes.

For a more realistic image, start off with a 960x1080 monochrome image (such as for 4:2:2 U or V), then scale to 960x540 (to simulate 4:2:0). Now scale back up to 960x1080. Information (vertical resolution) must have been lost in the process, but the rescaling will have interpolated the missing lines as well as it can.

Comparing this with the original 960x1080 shows how much INFORMATION got lost in the first downscaling process, as opposed to pixel sites. You're obviously going to lose half the pixel sites, but much less than half the information, and this is what is important to a coder. It exploits that any given pixel in a real image is likely to generally have a very similar value to it's neighbours - it's exploiting REDUNDANCY in the chroma signal. It's why you don't need twice the number of bits to code twice as many pixels.

Now all that is why I think any statement like "all 50Mbs gives you is 4:2:2" is a fallacy, albeit an oft-quoted one. But coming back to the real world, the question is "well OK, but even so, does it matter?" And hence on to the other points you and Simon bring up. (And which, by and large, I have a lot of agreement with.)

There may well be an argument for getting such as the BBC to accept the PMW350 as it stands - but then I can see a much bigger argument in getting Sony to upgrade it to 422 50Mbs. (Which I'm sure is technically possible at not much extra cost.) At which point all acceptance arguments disappear into the sunset.

[Simon Wyndham]

My feeling is that the 350 will get a sister model with 50Mb/s. XDCAM camcorders usually come in pairs (510-530, 330-350, 335-355, 700-800). A 50Mb/s version of the 350 wouldn't encroach on any other of the Sony camcorders.

Any such model would be closer to the 700 in pricing (or halfway between 350 and 700). But it wouldn't impact on the 700 because some simply want the disc workflow instead of solid state. Furthermore the performance of the 350 as it already stands, plus the use of a NanoFlash, means that if the 350 was to encroach and impact on the 700 then it has done already. So it could make sense for Sony to offer a model with 50Mb/s recording for a couple of grand more to save owners from putting an external device onto it. Something that they would do anyway even if there wasn't a sister model.

[Alister Chapman]

I doubt we will see a 50Mb/s PMW-350. It would compete directly with the 700 and Sony like to have distinct product lines in distinct markets. As I see it:

EX1/EX3 replaces DSR300 budget 1/2" DVCAM for low cost corporate.
PMW-350 replaces DSR-450/DSR570 for news and corporate sector.
PDW-700/F800 replaces Digibeta for general broadcast.

Certainly many people will prefer the optical disc workflow over solid state but many people, myself included bought in to XDCAM HD422 primarily because it is accepted for broadcast out of the box, pretty much universally. There are many PDW users desparately waiting for Sony to release the SxS adapter so they can switch to solid state. Lots of rumors as to what that will do, 50Mb/s maybe more?? it will almost certainly have HDSDi in.

[Mike Marriage]

...and I think that like DVCAM vs Digibeta, 2/3" XDCAM will become used for a huge number of broadcast shows, including primetime.

The nano flash is an interesting curve ball in the mix as well.

[Simon Wyndham]

I guess we'll have to wait and see.

[David Heath]

Originally Posted by Alister Chapman

I doubt we will see a 50Mb/s PMW-350.

PMW-350 replaces DSR-450/DSR570 for news and corporate sector.
PDW-700/F800 replaces Digibeta for general broadcast.

I tend to agree more with Simon. The above may be true as far as the first sentence goes (PMW350 for news and higher end corporate), but broadcasters are increasingly looking towards solid state for all acquisition. The 700/800 will be some peoples preference for a while to come, but increasingly I see people wanting XDCAM422 (or equivalent codec) natively in a solid state camera. I just don't see having to choose between solid state OR XDCAM422 a good sales ploy for long.

I also find it strange that Sony have considerably upgraded the EX1 - but not the EX3. My guess is there are more announcements to come, either upgrading the EX3, or discontinuing it in favour of a completely new model more nearly half way between the EX1R and the PMW350.

[Doug Jensen]

Originally Posted by Simon Wyndham

My feeling is that the 350 will get a sister model with 50Mb/s. XDCAM camcorders usually come in pairs (510-530, 330-350, 335-355, 700-800).

I disagree with that. When you look at pairs of cameras, they usually both have the same basic recording formats. What you get for a little extra $$ is bells and whistles and not a totally different HD recording mode. If there ever is a PMW-355, I predict it will only have a few extra features not found on the 350.

Perhaps a PMW-400 with 50Mb/s is more likely, but don't bet on that either.
I do agree that an EX3R or totally new camera must be in the works because there is a big gaping whole in the product line.

[Alister Chapman]

I've been told there are no plans for an EX3R in the near future and that the EX1R was introduced (in part) as a companion camera for the PMW-350 to give SD users of the 350 a solid state SD "B" camera that uses the same workflow.

But then I see Bob Ott saying that there are some surprises to come at NAB.

[Uli Mors]

Question is: Will the SXS cards have enough (writing) transfer speed (50mbit/s + headroom for safety or overcranking) to make a 50mbit recording?

There is a SXS recording unit for pdw700 in the line - but nobody knows exactly, if 50mbit recording is supported...

Uli

[David Heath]

Originally Posted by Alister Chapman

I've been told there are no plans for an EX3R in the near future.....

As I see it, there are effectively three options:

1/ Upgrade the EX3 in a similar way to the EX1.
2/ Do nothing - keep the range as EX1R, EX3, and PMW350.
3/ Discontinue the EX3, and replace it with a new model - maybe a shouldermount 1/2" SxS - cheaper than the PMW350.

You rule out option 1, and I don't think option 2 is sustainable in a marketing sense. How can anyone charge more for a "higher spec" model, when the one below it has a lot of desirable features it's missing?

Hence my money must strongly be on option 3.